Overclock.net › Forums › Industry News › Hardware News › [Various] Futuremark Releases 3DMark Time Spy DirectX 12 Benchmark
New Posts  All Forums:Forum Nav:

[Various] Futuremark Releases 3DMark Time Spy DirectX 12 Benchmark - Page 38

post #371 of 772
Since there is a benchmark thread on this, im going to take leave of this one. I am sick and tired of the constant back and forth in the news threads of red vs green. Have fun.
post #372 of 772
Here is the 980 vs 1070 in identical clock and memory speed. (at 1430 core/6804 mem). I think 960 vs 1060 at identical speeds would give most accurate results but this would do for now.
Its in Korean but you don't need the translation to read bench results. Consider 1070 has 1920 shaders vs 980's 2048, they both got 64 ROP.


http://www.hwbattle.com/bbs/board.php?bo_table=hottopic&wr_id=2243








Pascal does have slight improvement from Maxwell other than process node but doesn't look significant.
Edited by Kpjoslee - 7/15/16 at 6:36pm
My home PC
(16 items)
 
  
CPUMotherboardGraphicsRAM
AMD Threadripper 1950x Gigabyte Aorus X399 Gaming 7  MSI Geforce GTX 1080ti Gaming X G.Skill DDR4 3600 CL16 
Hard DriveHard DriveCoolingOS
Samsung Evo 840 500GB Samsung 960 Pro 500GB Noctua NH-U14S TR4 Windows 10 Pro 
MonitorMonitorKeyboardPower
Dell U2711 Samsung 55" 4k Corsair K70  EVGA SuperNova G2 1300W 
CaseMouseAudio
Corsair Carbide Air 540 Logitech G502 Denon AVR-X3300W 
  hide details  
Reply
My home PC
(16 items)
 
  
CPUMotherboardGraphicsRAM
AMD Threadripper 1950x Gigabyte Aorus X399 Gaming 7  MSI Geforce GTX 1080ti Gaming X G.Skill DDR4 3600 CL16 
Hard DriveHard DriveCoolingOS
Samsung Evo 840 500GB Samsung 960 Pro 500GB Noctua NH-U14S TR4 Windows 10 Pro 
MonitorMonitorKeyboardPower
Dell U2711 Samsung 55" 4k Corsair K70  EVGA SuperNova G2 1300W 
CaseMouseAudio
Corsair Carbide Air 540 Logitech G502 Denon AVR-X3300W 
  hide details  
Reply
post #373 of 772
Quote:
Originally Posted by Kpjoslee View Post

Warning: Spoiler! (Click to show)






Pascal does have slight improvement from Maxwell other than process node but doesn't look significant.

Would be nice if they used the same drivers for both Maxwell and Pascal...



Overall, similar performance per clock but much better performance per watt. Seems the performance gains were largely due to 16nm FinFET higher clocks.
Tsar Bomba
(12 items)
 
Fat Man
(13 items)
 
Little Boy
(29 items)
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K ASRock Fatal1ty Z170 Gaming-ITX/ac Crucial Ballistix Sport LT 32GB (2 x 16GB) DDR4... Samsung 850 EVO 500GB M.2-2280 SSD 
Hard DriveCoolingCoolingOS
Sandisk Ultra II 960GB 2.5" SSD Zalman CNPS8900 Quiet Noctua NF-F12 PWM Microsoft Windows 10 Pro 
KeyboardPowerCaseMouse
Corsair K95 Corsair SF 600W Silverstone SG05BB-LITE Corsair M65 
CPUMotherboardGraphicsGraphics
AMD A10-7850K GIGABYTE GA-F2A88X-UP4 XFX R9-290A-ENBC Radeon R9 290 XFX R9-290A-ENBC Radeon R9 290 
Hard DriveHard DriveCoolingCooling
Samsung 830 128GB Western Digital Blue 1TB XSPC Raystorm CPU/APU Waterblock - AMD XSPC D5 Photon 170 Reservoir/Pump Combo 
CoolingCoolingCoolingPower
XSPC AX360 Radiator XSPC AX240 Radiator 2x XSPC Razor R9 290X / 290 - Full Cover Water ... EVGA Supernova 1000 P2 
Case
Corsair Carbide Series Air 540 Black  
CPUMotherboardGraphicsGraphics
Intel Core i7-4770K  Asus Maximus VI Extreme Sapphire R9 290 unlocked to 290X Sapphire R9 290 unlocked to 290X 
GraphicsGraphicsRAMHard Drive
Sapphire R9 290 unlocked to 290X Sapphire R9 290 unlocked to 290X G.SKILL Sniper Series 32GB (4 x 8GB) DDR3 2400 2x SAMSUNG 840 EVO 250GB 
Hard DriveOptical DriveCoolingCooling
2x TOSHIBA PH3300U-1I72 3TB Asus BW-12B1ST/BLK/G/AS Blu-Ray/DVD/CD Writer 29x Noctua NF-F12 PWM 2x Swiftech MCP655 + Bitspower Dual D5 Mod Top 
CoolingCoolingCoolingCooling
Phobya Balancer 450 Silver Nickel Reservoir 2x Alphacool NexXxoS XT45 480mm Radiator Alphacool NexXxoS XT45 360mm Radiator Alphacool NexXxoS XT45 240mm Radiator 
CoolingCoolingCoolingCooling
Alphacool NexXxoS ST30 240mm Radiator Alphacool NexXxoS XT45 120mm Radiator 4x Koolance VID-AR290X Radeon VGA Liquid Coolin... EK Supreme LTX Intel CPU Liquid Cooling Block -... 
OSMonitorMonitorMonitor
Microsoft Windows 8.1 Pro Asus VS238H-P 23.0" Asus VS238H-P 23.0" Asus VS238H-P 23.0" 
KeyboardPowerCaseMouse
Corsair Vengeance K95 ENERMAX Maxrevo EMR1500EWT 1500W Corsair Obsidian Series 900D Razer DeathAdder 2013 
Audio
Astro Gaming A50 Circumaural Headset - Black 
  hide details  
Reply
Tsar Bomba
(12 items)
 
Fat Man
(13 items)
 
Little Boy
(29 items)
 
CPUMotherboardRAMHard Drive
Intel Core i7-6700K ASRock Fatal1ty Z170 Gaming-ITX/ac Crucial Ballistix Sport LT 32GB (2 x 16GB) DDR4... Samsung 850 EVO 500GB M.2-2280 SSD 
Hard DriveCoolingCoolingOS
Sandisk Ultra II 960GB 2.5" SSD Zalman CNPS8900 Quiet Noctua NF-F12 PWM Microsoft Windows 10 Pro 
KeyboardPowerCaseMouse
Corsair K95 Corsair SF 600W Silverstone SG05BB-LITE Corsair M65 
CPUMotherboardGraphicsGraphics
AMD A10-7850K GIGABYTE GA-F2A88X-UP4 XFX R9-290A-ENBC Radeon R9 290 XFX R9-290A-ENBC Radeon R9 290 
Hard DriveHard DriveCoolingCooling
Samsung 830 128GB Western Digital Blue 1TB XSPC Raystorm CPU/APU Waterblock - AMD XSPC D5 Photon 170 Reservoir/Pump Combo 
CoolingCoolingCoolingPower
XSPC AX360 Radiator XSPC AX240 Radiator 2x XSPC Razor R9 290X / 290 - Full Cover Water ... EVGA Supernova 1000 P2 
Case
Corsair Carbide Series Air 540 Black  
CPUMotherboardGraphicsGraphics
Intel Core i7-4770K  Asus Maximus VI Extreme Sapphire R9 290 unlocked to 290X Sapphire R9 290 unlocked to 290X 
GraphicsGraphicsRAMHard Drive
Sapphire R9 290 unlocked to 290X Sapphire R9 290 unlocked to 290X G.SKILL Sniper Series 32GB (4 x 8GB) DDR3 2400 2x SAMSUNG 840 EVO 250GB 
Hard DriveOptical DriveCoolingCooling
2x TOSHIBA PH3300U-1I72 3TB Asus BW-12B1ST/BLK/G/AS Blu-Ray/DVD/CD Writer 29x Noctua NF-F12 PWM 2x Swiftech MCP655 + Bitspower Dual D5 Mod Top 
CoolingCoolingCoolingCooling
Phobya Balancer 450 Silver Nickel Reservoir 2x Alphacool NexXxoS XT45 480mm Radiator Alphacool NexXxoS XT45 360mm Radiator Alphacool NexXxoS XT45 240mm Radiator 
CoolingCoolingCoolingCooling
Alphacool NexXxoS ST30 240mm Radiator Alphacool NexXxoS XT45 120mm Radiator 4x Koolance VID-AR290X Radeon VGA Liquid Coolin... EK Supreme LTX Intel CPU Liquid Cooling Block -... 
OSMonitorMonitorMonitor
Microsoft Windows 8.1 Pro Asus VS238H-P 23.0" Asus VS238H-P 23.0" Asus VS238H-P 23.0" 
KeyboardPowerCaseMouse
Corsair Vengeance K95 ENERMAX Maxrevo EMR1500EWT 1500W Corsair Obsidian Series 900D Razer DeathAdder 2013 
Audio
Astro Gaming A50 Circumaural Headset - Black 
  hide details  
Reply
post #374 of 772
Quote:
Originally Posted by ChevChelios View Post

BF1 full Ultra 1440p DX11 already gives ~90 avf fps on alpha right now on a 6700K + GTX1080

it can only get better with release polish and a driver wink.gif

God, does your shilling no no bounds? Nobody is suggesting that a $700 Nvidia flagship GPU won't perform adequately in BF1 regardless of API, that's to be expected. The question is how lower tier Nvidia cards and older generation cards will do. The fact is that the 480 is $200 and is guaranteed to do better in DX12 games like BF1 (and all AAA games moving forward) than its competitor the 1060 which is extremely relevant in the most popular bracket of GPU sales. What is so hard for you to understand about that? I mean, I know you have spent months attempting to downplay DX12 at every opportunity but you have to face facts at this point that DX11 is basically dead right now when it comes to the big titles coming up and with Win10 ubiquity that will only become more and more true over the next 12 months, so if you have an older card like a 290X you will be in way better shape than those of us on equivalent older HW from Nvidia (like my Titans which will become almost irrelevant with little DX12 support).
post #375 of 772
Quote:
Originally Posted by Kielon View Post

As a consumer, I expect to have close to optimal performance from day one. I will not buy a Tesla car if I live in a country with no superchargers in place or I will not buy 4K TV if there is no 4K content. The same rule applies to DX11 and DX12 content.

So AMD sucks at DX11 but excels at DX12 and they therefore suck, but Nvidia sucks at DX12 and excels at DX11 so they are somehow amazing? Neither brand excels at BOTH yet as usual only AMD is considered crap. Nice logic...
post #376 of 772
Quote:
Originally Posted by dagget3450 View Post

It's not going to be much help in some areas on checking performance for sure. In Doom on my FuryX i can go from NOAA to TSSAA8X with 0 fps loss. 3dmark doesn't appear to have anything for using AA which would be a nice way to see if there is hit on performance. Also will not be able to compare green vs red in the same area.

Anyways, i keep seeing this bench is able to use multi adapter but i don't see an option for it? Linke/unlinked gpu? Anyone have anything information on this?

Well... when you consider that you program Multi-Adapter the same way you program for Asynchronous Compute + Graphics under DX12 then you can pretty much conclude that Time Spy only makes a concurrent usage of Asynchronous Compute not a parallel usage. If the usage were parallel then using multi-adapters would result in very good scaling as in AotS for example.

A few things worth considering...


Quote:
I'm talking OS level pre-emption. What you are talking about on GCN is more like hyperthreading. You have the equivalent of multiple threads being executed at hardware level, but you can't really pre-empt a specific thread. That is, the OS can't stop a task on a specific GPU queue, then switch something, then switch the old job out. It has to let the current GPU job continue on that queue. For Multi-GPU, we actually submit our command lists in sections so that the OS has a chance to swap in our present during the middle of a frame so we can flip the backbuffer. I suspect the way D3D12 works is that the GPU creates a CPU interrupt when a queue signals something, then the Windows Kernal submits the next task to the GPU for completion if there is a ->Wait sitting on the queue. You could see that this round trip to the CPU back to the GPU could slow it down a tiny bit, because AFAIK the GPU is required to flush before continuing. You can also see why MS made this setup - because now dozens of applications could actually be giving work to the GPU, and theoretically the OS could schedule them all. Windows is more then a game OS, afterall.

Yes, you are correct about what happens with async off. We submit the tasks to the universal queue and then don't bother submitting fences and signals because we don't need.

GCN has multiple queues, which can execute in parallel, but that's not the same thing as pre-emption. Time slicing on GCN occurs on a hardware scheduler, and GCN can actually synchronize at a very fine grain. The best way to think about it the ACEs basically look like extra GPUs. However, D3D12 synchronization primitives are at an OS level, the hardware doesn't actually see them. I think they did it this way so that it's more unified across different hardware types. AFAIK, Fences and signals aren't actually directly visible to the driver. TLDD: GCN can actually do much more fine grained synchronization then D3D12 allows right now.

All of this can be seen by an ETW trace of Ashes, you can actually see tiny GPU stalls on GCN where the signals happen. I'd guess we are losing 2-3% perf because of it, but that's the price you pay for being on a multi-tasking OS.
Quote:
The tasks have fences and signals on them, and I believe as part of the D3D12 specification a fence ends up flushing the GPU, which could mean a 100 us stall or so. TLDR is that adding a command to a compute queue or any queue , or rather the act of synchronizing it will have a tiny bit of overhead. Thus, if the hardware doesn't have some intrinsic gain from doing it parallel, you'll likely end up with a tiny loss. Even an architecture that can do them in parrelel will likely lose a little bit. It's just that the net gain is more then the loss.
Quote:
I'm not sure you understand the nature of this feature. There are 2 main types of tasks for a GPU, graphics and compute. D3D12 exposes main 2 queue types, a universal queue (compute and graphics), and a compute queue. For Ashes, use of this feature involves taking compute jobs which are already part of the frame and marking them up in a way such that hardware is free to coexecute it with other work. Hopefully, this is a relatively straightfoward tasks. No additional compute tasks were created to exploit async compute. It is merely moving work that already exists so that it can run more optimally. That is, if async compute was not present, the work would be added to the universal queue rather than the compute queue. The work still has to be done, however.

The best way to think about it is that the scene that is rendered remains (virtually) unchanged. In D3D12 the work items are simply arranged and marked in a manner that allows parallel execution. Thus, not using it when you could is seems very close to intentionally sandbagging performance.

The difference is pretty staggering between the Time Fly Demo and the games currently out. The games pretty much execute Asynchronous Compute tasks in parallel while Time Fly executes them concurrently. It is thus not the same Asynchronous Compute. Concurrent execution simply fill up gaps in the pipeline whereas parallel executions cut down the execution times of parallel running tasks. When executing in parallel... you can execute both a compute and a graphics task at the same time but with a synchronization point between the two. If Time Fly where using this method then we would see a performance loss for Maxwell GPUs... and we do not see that. Why a performance loss? Because the synchronization requires a flush (which takes time on Maxwell but which was improved on Pascal).

It is discussed in nVIDIAs GTX 1080 whitepaper here...


So instead... what we have are concurrent executions which do not cause this issue with Maxwell as Graphics and Compute tasks are not synchronized. You also avert a flush because you are not using any sync points. This is why the Maxwell series do not suffer a performance loss under the Time fly Demo when running Async compute.

So yeah... 3DMark Time fly is not the same Asynchronous Compute used in all of the games out currently. thumb.gif It could however indicate the type of Async compute that Bethesda will be implementing into Doom for nVIDIA hardware.
Edited by Mahigan - 7/15/16 at 7:57pm
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
post #377 of 772
Quote:
Originally Posted by Majin SSJ Eric View Post

God, does your shilling no no bounds? Nobody is suggesting that a $700 Nvidia flagship GPU won't perform adequately in BF1 regardless of API, that's to be expected. The question is how lower tier Nvidia cards and older generation cards will do. The fact is that the 480 is $200 and is guaranteed to do better in DX12 games like BF1 (and all AAA games moving forward) than its competitor the 1060 which is extremely relevant in the most popular bracket of GPU sales. What is so hard for you to understand about that? I mean, I know you have spent months attempting to downplay DX12 at every opportunity but you have to face facts at this point that DX11 is basically dead right now when it comes to the big titles coming up and with Win10 ubiquity that will only become more and more true over the next 12 months, so if you have an older card like a 290X you will be in way better shape than those of us on equivalent older HW from Nvidia (like my Titans which will become almost irrelevant with little DX12 support).

AMD is not better at DX12 then Nvidia. It's just that AMD has hardware ASync. If a game does not have ASync but is DX12 then all DX12 does for AMD is bring it to Nvidia level. This was a shortcoming of AMD and nothing more. We pained for DX11 performance. AMD had to use stronger GPU to match Nvidia in DX11. Now that they are in the same playing field with DX12/VulKan you see that extra performance.
Ishimura
(21 items)
 
Silent Knight
(13 items)
 
 
CPUMotherboardGraphicsRAM
Intel Core i7 3770K @ 4.6GHz ASRock Z77E-ITX eVGA GTX 1080 Ti Hybrid AVEXIR Blitz 1.1 16GB DDR3-2400MHz CL10 
Hard DriveHard DriveCoolingCooling
SanDisk Ultra II 960GB Toshiba X300 5TB Corsair H100i GTX eVGA Hybrid Water Cooler  
CoolingOSMonitorKeyboard
4x GentleTyphoon AP-15 Windows 10 Pro 64-Bit Philips Brilliance BDM4065UC 4K Razer BlackWidow Chroma  
PowerCaseMouseMouse Pad
eVGA SuperNOVA 750 G3 Define Nano S Logitech G502 Proteus Core PECHAM Gaming Mouse Pad XX-Large 
AudioAudioAudioAudio
Audioengine D1 DAC Mackie CR Series CR3 Audio-Technica ATH-M50 Sennheiser HD 598 
Audio
Sony XB950BT 
CPUMotherboardGraphicsRAM
AMD Phenom II X4 955 @ 4.2GHz ASUS M4A79XTD EVO AMD Radeon HD 7970 3GB @ 1200/1500 2x 4GB G.SKILL Ripjaws X DDR3-1600 
Hard DriveHard DriveHard DriveCooling
OCZ Agility 3 60GB WD Caviar Green 1.5TB 2 x Seagate Barracuda 2TB XSPC Raystorm 
CoolingCoolingOSPower
EK-FC7970 XSPC RS360 Windows 10 Pro 64-Bit Corsair TX750 
Case
NZXT Switch 810  
  hide details  
Reply
Ishimura
(21 items)
 
Silent Knight
(13 items)
 
 
CPUMotherboardGraphicsRAM
Intel Core i7 3770K @ 4.6GHz ASRock Z77E-ITX eVGA GTX 1080 Ti Hybrid AVEXIR Blitz 1.1 16GB DDR3-2400MHz CL10 
Hard DriveHard DriveCoolingCooling
SanDisk Ultra II 960GB Toshiba X300 5TB Corsair H100i GTX eVGA Hybrid Water Cooler  
CoolingOSMonitorKeyboard
4x GentleTyphoon AP-15 Windows 10 Pro 64-Bit Philips Brilliance BDM4065UC 4K Razer BlackWidow Chroma  
PowerCaseMouseMouse Pad
eVGA SuperNOVA 750 G3 Define Nano S Logitech G502 Proteus Core PECHAM Gaming Mouse Pad XX-Large 
AudioAudioAudioAudio
Audioengine D1 DAC Mackie CR Series CR3 Audio-Technica ATH-M50 Sennheiser HD 598 
Audio
Sony XB950BT 
CPUMotherboardGraphicsRAM
AMD Phenom II X4 955 @ 4.2GHz ASUS M4A79XTD EVO AMD Radeon HD 7970 3GB @ 1200/1500 2x 4GB G.SKILL Ripjaws X DDR3-1600 
Hard DriveHard DriveHard DriveCooling
OCZ Agility 3 60GB WD Caviar Green 1.5TB 2 x Seagate Barracuda 2TB XSPC Raystorm 
CoolingCoolingOSPower
EK-FC7970 XSPC RS360 Windows 10 Pro 64-Bit Corsair TX750 
Case
NZXT Switch 810  
  hide details  
Reply
post #378 of 772
Quote:
Originally Posted by magnek View Post

I know how you feel.. I'm still a bit salty that I can't get my 980 Ti to 1600 despite having way more than enough cooling headroom (420+360+240) simply because voltage scaling is now no longer a thing under ambient settings.
Yep Fire Strike definitely does that to your card. Although Heaven consistently pushes a high sustained power draw which is why it's awesome for stability testing.

Good old OG Titans still rule as far as scaling from voltage is concerned! I can increase my clocks by 11% due to voltage increase. Of course the arch is just so old now that it doesn't matter as stock Maxwell and Pascal cards are much faster overall. Still, its nice that my W/C has a purpose with these old Kepler cards...
post #379 of 772
Quote:
Originally Posted by Remij View Post

LOL who really cares what they do to get the performance they do as long as image quality isn't affected?

AMD could put all the compute engines, anti-aliasing engines, effect engines with 10+ hardware schedulers and fancy parallelization it wants. If it still gets beat by brute force Nvidia hardware, why does it matter? biggrin.gif

Maybe we should. interesting to know if difference between tech actually makes a difference. But then again if your just into Brute Force then whocares.
post #380 of 772
Quote:
Originally Posted by ht_addict View Post

Maybe we should. interesting to know if difference between tech actually makes a difference. But then again if your just into Brute Force then whocares.

Eh, performance is performance. The problem with brute force is cost. These Titan X's that are brute forcing their performance these days did cost in excess of $1000, lets not forget...
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Various] Futuremark Releases 3DMark Time Spy DirectX 12 Benchmark